http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
단일채널 수면뇌파 분석을 위한 컨볼루션 신경망 모델 최적화
유현수(Hyunsoo Yu),한승우(Seungwoo Han),박철수(Cheolsoo Park) 대한전자공학회 2020 대한전자공학회 학술대회 Vol.2020 No.8
Sleep is very important for staying in a healthy condition. Research on sleep has been conducted intensively in recent years. Sleep scoring has been conducted by sleep experts with polysomnography which is complex process. Estimating sleep patterns with single channel EEG(Electroencephalogram) is one of the methods to automate the process for sleep analysis. An CNN (Convolutional Neural Network) model was used in sleep scoring. Additionally, the CNN model was optimized with Bayesian Optimization and Hyperband (BOHB). The accuracy from the optimized model was higher than the accuracy from empirically designed model.
시계열 데이터 분석을 위한 컨볼루션 신경망 기반의 Empirical Mode Decomposition
유현수(Hyunsoo Yu),백수환(Suwhan Baek),박철수(Cheolsoo Park) 대한전자공학회 2021 대한전자공학회 학술대회 Vol.2021 No.6
EMD (Empirical Mode Decomposition) was proposed as a method for analyzing nonlinear and nonstationary data. In this study, an attempt was made to implement EMD with deep learning. The model used was a simple CNN (Convolutional Neural Network), and after prediction, it was shown that the prediction result and the true signal showed a similar tendency especially in the low-order IMF (Intrinsic Mode Function).
Single channel EEG의 Attention을 통한 수면 단계 추정 의사결정지원시스템
백수환(Suwhan Baek),백재우(Jaewu Baek),유현수(Hyunsoo Yu),박철수(Cheolsoo Park) 대한전자공학회 2021 대한전자공학회 학술대회 Vol.2021 No.6
This paper propose classification model for AASM(American Academy of Sleep Medicine) based five level sleep staging(Wake, Rem, N1, N2, N3) by single channel EEG signal(F3). n particular, for explainable sleep phase estimation models unlike conventional models, we proceed with the analysis of signals via BEMD(Bidimensional Empirical Mode Decomposition), and prove that each sleep label was learned from which key signal components through back-analysis of the model through attention mechanisms. The cross-validation results of 425 subjects" data with signal analysis for Single Channel EEG showed an average of F1-Score 70.232 (±1.769) average accuracy of 83.15% (±1.35%) and we propose a explainable model for which electroencephalogram components were more effective in estimating each sleep stage.